2nd International ICST Workshop on Network Coding, Theory, and Applications

Research Article

An Analysis of Finite-Memory Random Linear Coding on Packet Streams

  • @INPROCEEDINGS{10.1109/WIOPT.2006.1666502,
        author={Desmond S.  Lun and Muriel  Medard and Payam  Pakzad and Christina  Fragouli and Ralf  Koetter},
        title={An Analysis of Finite-Memory Random Linear Coding on Packet Streams},
        proceedings={2nd International ICST Workshop on Network Coding, Theory, and Applications},
        publisher={IEEE},
        proceedings_a={NETCOD},
        year={2006},
        month={8},
        keywords={},
        doi={10.1109/WIOPT.2006.1666502}
    }
    
  • Desmond S. Lun
    Muriel Medard
    Payam Pakzad
    Christina Fragouli
    Ralf Koetter
    Year: 2006
    An Analysis of Finite-Memory Random Linear Coding on Packet Streams
    NETCOD
    IEEE
    DOI: 10.1109/WIOPT.2006.1666502
Desmond S. Lun1,*, Muriel Medard1,*, Payam Pakzad2,*, Christina Fragouli2,*, Ralf Koetter3,*
  • 1: Laboratory for Information and Decision Systems, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
  • 2: Laboratoire d’algorithmique, Ecole Polytechnique Federale de Lausanne, CH-1015 Lausanne, Switzerland
  • 3: Coordinated Science Laboratory, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA
*Contact email: dslun@mit.edu, medard@mit.edu, payam.pakzad@epfl.ch, christina.fragouli@epfl.ch, koetter@uiuc.edu

Abstract

We consider the following packet coding scheme: The coding node has a fixed, finite memory in which it stores packets formed from an incoming packet stream, and it sends packets formed from random linear combinations of its memory contents. We analyze the scheme in two settings: as a self-contained component in a network providing reliability on a single link, and as a component employed at intermediate nodes in a block-coded end-to-end connection. We believe that the scheme is a good alternative to automatic repeat request when feedback is too slow, too unreliable, or too difficult to implement.